The Duffing Equation

The Duffing equation is an externally forced and damped oscillator equation. It is given by
.

We will investigate the equation by turning on a few constants at a time. We begin with simple harmonic motion:
0.

clear
alpha = 1; delta = 0; beta = 0; % SHM
duffing = @(t,y) [y(2); -delta*y(2)-alpha*y(1)-beta*y(1).^3];
tspan = [0 30];
y0 = [0 0.01]';
[t,y] = ode45(duffing,tspan,y0);
plot(t,y)
legend({'y','dy/dt'})
xlabel('t')
ylabel('y')
title('Simple Harmonic Motion')

figure
plot(y(:,1),y(:,2))
xlabel('x')

1
 ylabel('y')
title('SHM Trajectory')
axis equal

Now add damping,

clear
alpha = 1; delta = 0.1; beta = 0; % damped SHM
duffing = @(t,y) [y(2); -delta*y(2)-alpha*y(1)-beta*y(1).^3];
tspan = [0 30]; % tspan = 0:.1:30; % Makes solution smoother
y0 = [0 0.1]';
[t,y] = ode45(duffing,tspan,y0);
plot(t,y)
legend({'y','dy/dt'})
xlabel('t')
ylabel('y')
title('Damped Simple Harmonic Motion')

2
figure
plot(y(:,1),y(:,2))
xlabel('x')
ylabel('y')
title('Damped SHM Trajectory')
axis equal

3
Now, consider the undamped Duffing equation As a first order system, we have

. Note the equilibrium points are .

clear
alpha = -1; delta = 0; beta = 1; % Undamped Duffing Equation
duffing = @(t,y) [y(2); -delta*y(2)-alpha*y(1)-beta*y(1).^3];
tspan = [0 10];
y0 = [1 1]';
[t,y] = ode45(duffing,tspan,y0);
plot(t,y)
legend({'y','dy/dt'})
xlabel('t')
ylabel('y')
title('Undamped, Unforced Duffing Equation')

4
figure
plot(y(:,1),y(:,2))
xlabel('x')
ylabel('y')
title('Duffing Equation Trajectory')
axis equal

5
Its useful to look at the phase plot using the quiver function. N is used to normalize the arrows in the direction
field. The first plot shows one trajectory.

[X, Y] = meshgrid(-1.5:0.25:1.5, -1.5:0.25:1.5);

U = Y; V=-delta*Y-alpha*X-beta*X.^3;
N=sqrt(U.^2+V.^2);
quiver(X, Y, U./N, V./N)

hold on
plot(y(:,1),y(:,2),'r')
hold off
xlabel('x')
ylabel('y')
title('Duffing Equation Quiver Plot')

6
The following shows how to add several trajectories.

hold on
for j=1:5
y0 = [1 j/5]';
[t,y] = ode45(duffing,tspan,y0);
plot(y(:,1),y(:,2),'r')
end
hold off

7
Now, we add back some damping. Changing the initial condition, y0 = [1 1]'; will reveal different types of
behavior.

clear
alpha = -1; delta = 0.1; beta = 1; % Damped Duffing Equation
duffing = @(t,y) [y(2); -delta*y(2)-alpha*y(1)-beta*y(1).^3];
tspan = [0 30];
y0 = [1 1]';
[t,y] = ode45(duffing,tspan,y0);
plot(t,y)
legend({'y','dy/dt'})
xlabel('t')
ylabel('y')
title('Damped Duffing Equation')

8
figure
plot(y(:,1),y(:,2))
xlabel('x')
ylabel('y')
title('Duffing Equation Trajectory')
axis equal

9
Here we add the forcing.

clear
alpha = -1; delta = 0.1; beta = 1; gamma = .1; omega=1.4; % Forced DE
duffing = @(t,y) [y(2); -delta*y(2)-alpha*y(1)-beta*y(1).^3+gamma*cos(omega*t)];
tspan = 0:0.1:200;
y0 = [0,0]';
[t,y] = ode45(duffing,tspan,y0);
plot(t,y)
legend({'y','dy/dt'})
xlabel('t')
ylabel('y')
title('Damped, forced Duffing Equation')

10
figure
plot(y(:,1),y(:,2))
xlabel('x')
ylabel('y')
title('Duffing Equation Trajectory')
axis equal

11
We can examine the tail of the trajectory and notice a periodic behavior with period

% Plot tail of trajectory

figure
L=length(y(:,1));
Npts=45; % Npts+1 points of the tail
plot(y(L-Npts:L,1),y(L-Npts:L,2))
%plot(y(L-44:L,1),y(L-44:L,2)) % roughly 45*.1=period
xlabel('x')
ylabel('y')
title(['Tail of Trajectory \gamma = ',num2str(gamma)])
axis equal

12
 T=2*pi/omega

T = 4.4880

If we change the amplitude to then we notice period doubling.

clear % Change gamma to 0.318 - Period Doubling to 2 cycles

alpha = -1; delta = 0.1; beta = 1; gamma = .318; omega=1.4; % Forced DE
duffing = @(t,y) [y(2); -delta*y(2)-alpha*y(1)-beta*y(1).^3+gamma*cos(omega*t)];
NT = 800;
tspan = 0:0.1:NT;
y0 = [0,0]';
[t,y] = ode45(duffing,tspan,y0);
plot(t,y)
legend({'y','dy/dt'})
xlabel('t')
ylabel('y')
title('Undamped, Unforced Duffing Equation')

13
figure
plot(y(:,1),y(:,2))
xlabel('x')
ylabel('y')
title('Duffing Equation Trajectory')
axis equal

14
% Plot tail of trajectory
figure
L=length(y(:,1));
Npts=180; % Npts+1 points of the tail
%plot(y(L-Npts:L,1),y(L-Npts:L,2))
plot(y(L-90:L,1),y(L-90:L,2)) % 90*.1=period - double before
xlabel('x')
ylabel('y')
title(['Tail of Trajectory \gamma = ',num2str(gamma)])
axis equal

15
 % Try gamma = 0.338 but 0.35 is chaotic region

For we get four cycles, or period 4.

clear % Change gamma to 0.338 - Period Doubling to 4 cycles

alpha = -1; delta = 0.1; beta = 1; gamma = .338; omega=1.4; % Forced DE
duffing = @(t,y) [y(2); -delta*y(2)-alpha*y(1)-beta*y(1).^3+gamma*cos(omega*t)];
NT = 2000;
tspan = 0:0.1:NT;
y0 = [0,0]';
[t,y] = ode45(duffing,tspan,y0);

figure
L=length(y(:,1));
Npts=180; % Npts+1 points of the tail
plot(y(L-Npts:L,1),y(L-Npts:L,2))
xlabel('x')
ylabel('y')
title(['Tail of Trajectory \gamma = ',num2str(gamma)])
axis equal

16
Can we find period 8? Maybe near

clear % - Period Doubling to more cycles?

alpha = -1; delta = 0.1; beta = 1; gamma = .3431; omega=1.4; % Forced DE
duffing = @(t,y) [y(2); -delta*y(2)-alpha*y(1)-beta*y(1).^3+gamma*cos(omega*t)];
NT = 6400;
tspan = 0:0.1:NT;
y0 = [0,0]';
[t,y] = ode45(duffing,tspan,y0);

figure
L=length(y(:,1));
Npts=360; % Npts+1 points of the tail
plot(y(L-Npts:L,1),y(L-Npts:L,2))
xlabel('x')
ylabel('y')
title(['Tail of Trajectory \gamma = ',num2str(gamma)])
axis equal

17
Possibly for there is only chaos.

clear % - Chaos?
alpha = -1; delta = 0.1; beta = 1; gamma = .35; omega=1.4; % Forced DE
duffing = @(t,y) [y(2); -delta*y(2)-alpha*y(1)-beta*y(1).^3+gamma*cos(omega*t)];
NT = 10000;
tspan = 0:0.1:NT;
y0 = [0,0]';
[t,y] = ode45(duffing,tspan,y0);

figure
L=length(y(:,1));
Npts=NT; % Npts+1 points of the tail
plot(y(L-Npts:L,1),y(L-Npts:L,2))
xlabel('x')
ylabel('y')
title(['Tail of Trajectory \gamma = ',num2str(gamma)])
axis equal

18
Let's look at a Poincaré Surface of Section for .

% Strange Attractor
clear
alpha = -1; delta = 0.1; beta = 1; gamma = .35; omega=1.4;
%alpha = -1; delta = 0.25; beta = 1; gamma = .40; omega=1; % Standard Shape
%alpha = -1; delta = 0.2; beta = 1; gamma = .30; omega=1; % Standard Shape

duffing = @(t,y) [y(2); -delta*y(2)-alpha*y(1)-beta*y(1).^3+gamma*cos(omega*t)];

NT = 6000; % or 2000 for gamma = 0.338
period=2*pi/omega;
dt=period/10;
tspan = 0:dt:20000;
y0 = [0,0]';
[t,y] = ode45(duffing,tspan,y0);

P_x = y(mod(t, period) < 1e-6, 1);

P_y = y(mod(t, period) < 1e-6, 2);

figure
scatter(P_x,P_y,'.')
xlabel('Position');
ylabel('Velocity');
title('Poincaré Surface of Section');

19
Here we add more initial conditions and find the same attractor. In fact, it is a strange attractor.

% Add more ICs

figure
scatter(P_x,P_y,'.')
xlabel('Position');
ylabel('Velocity');
title('Poincaré Surface of Section');
hold on
for j=1:50
y0 = [j/50,0]';
[t,y] = ode45(duffing,tspan,y0);
P_x = y(mod(t, period) < 1e-6, 1);
P_y = y(mod(t, period) < 1e-6, 2);
L=length(P_y);
Npts=180;
% scatter(P_x,P_y,'.r')
scatter(P_x(L-Npts:L),P_y(L-Npts:L),'.r')
end
hold off

20
A more standard strange attractor for the Duffing equation is found using a unit amplidute. Here are two
possibles cases with varying parameters.

% Strange Attractor
clear
alpha = -1; delta = 0.25; beta = 1; gamma = .40; omega=1; % Standard Shape
%alpha = -1; delta = 0.2; beta = 1; gamma = .30; omega=1; % Standard Shape

duffing = @(t,y) [y(2); -delta*y(2)-alpha*y(1)-beta*y(1).^3+gamma*cos(omega*t)];

NT = 6000; % or 2000 for gamma = 0.338
period=2*pi/omega;
dt=period/10;
tspan = 0:dt:20000;
y0 = [0,0]';
[t,y] = ode45(duffing,tspan,y0);

P_x = y(mod(t, period) < 1e-6, 1);

P_y = y(mod(t, period) < 1e-6, 2);

figure
scatter(P_x,P_y,'.')
xlabel('Position');
ylabel('Velocity');
title('Poincaré Surface of Section');

21
22